CN103064954B - Search based on entity and parsing - Google Patents

Abstract

Provide for by utilizing entity type sign to provide the system of abundant search experience, method and computer-readable storage medium.Search inquiry is received and entity type is determined for described inquiry.Then Web search for described inquiry is performed for the predetermined list being defined as fixed entity type is authoritative source.The result of the specific search in source is filtered based on described entity type, it is provided that the filtered list of result, and described result is respectively relevant with the entity of described entity type being defined as being associated with described inquiry.Described filtered list then by with resolve entity list compared with to determine that, by the equivalent entities of the different identifing source searched, described equivalent entities is combined into single potential Search Results.The Search Results of described merging is then based on rank value and is sorted.Then at least some of of described result can be presented to user.

Description

Search based on entity and parsing

Background technology

Given search inquiry character string, Web search engine returns the list of hyperlink traditionally, and described hyperlink is linked to be considered the page on the Web relevant to input search inquiry upon selection.Recently, search engine results page generally the most also includes more rich content via vertical information domain name.Along with this trend continues, search will converge to a bit, wherein, index and the retrieval of information are not performed only with respect to Web page, but also be performed relative to the such as application (such as, from application market), film, TV programme, people, famous person, event, city, restaurant, arenas, company etc..In order to emerge entity, search engine must crawl (crawl) multiple destructuring Web page and/or subscribe to the structuring summary about particular entity type, resolve the example of the entity across this multi-source data, and emerge the expression of described (merger) entity when the intention of user refers to described entity and/or its entity type.That execution operates as follows it is required that more complicated with the complication that index and searching entities are associated, described operation is: retrieve entity based on approximate description；The broad set some of which of retrieval entity directly may not be described by inquiry string；The metadata about it is retrieved in description based on the entity in unfashionable source from popular source；Usually combine feature and the ranking of the indexed entity across multiple sources；Entity is performed facet search；And become compound entirety to perform to integrate search generally by by the information integration from multiple webpages.

The prior art solution of entity search problem can be classified in two methods, and each method suffers from respective shortcoming.Initially, it is exclusively used in the vertical vertical engines results page (VERP) of single piece of information usually according to the index comprising primary entity attribute, the collection of the entity (such as, movie property) of single type to be scanned for.Such solution can lose efficacy in the inquiry provide semantically related text or the ambiguous description not appearing in index and (such as, inquire about " movie With a sinking boat starring DiCaprio (film of the shipwreck that DiCaprio acts the leading role) " film " Titanic (Titanic) " may not be returned, or inquiry " Batman (batman) " may not return film " The Dark Knight (dark knight) ").The method use Web search that the second is general, Web search has the advantage of the large-scale index of the relational language that make use of Web link structure and Anchor Text, and it includes powerful intention analysis, and employs automatic spelling correction.The shortcoming of the method is if the unused entity of the indexed page resolves, then the abundant content provided by VERP may never be emerged.And, because the indexed page is for being resolved each other, even if abundant content is retrieved, the large result of the example being linked to identical primary entity may be retrieved together, has slackened the multiformity of result.

Summary of the invention

Present invention is provided to introduce in simplified form the selection of concept, and these concepts are further described in following detailed description of the invention.Present invention is both not intended to identify key feature or the essential feature of theme required for protection, is also not intended to the scope being used to assist in determining theme required for protection.

Embodiments of the invention relate to being particularly useful for integrating the advantage of vertical search and the advantage of Web search to utilize entity type to characterize the system of search experience, method and the computer-readable storage medium providing abundant.Search inquiry is received and entity type is determined for described inquiry.Then Web search for described inquiry is performed for the predetermined list being defined as fixed entity type is authoritative source.Although certain source is probably authority's to described entity type, but it can also include the information for other entity type, and similarly, identifies the result relevant to multiple entity types.Therefore, the result of the specific search in source is filtered based on described entity type, it is provided that the filtered list of result, described result respectively be defined as relevant with the entity of the entity type that described inquiry is associated.Described filtered list then by with resolve entity list compared with to determine the equivalent entities identified by the different sources searched, described equivalent entities is combined into single potential Search Results.The Search Results of described merging is then based on rank value and is sorted.The described rank value of the entity being assigned to merging is the total rank value calculated from the single rank value provided for the entity from not homology.Described result be then presented to user at least partially.

Accompanying drawing explanation

The present invention is illustrated as an example and is not limited to accompanying drawing, wherein, and the unit that identical reference instruction is similar, and wherein:

Fig. 1 is suitable for the block chart of the exemplary calculated environment used in implementing embodiments of the invention；

Fig. 2 is the block chart of the wherein Exemplary computing systems that embodiments of the invention can be utilized；

The schematic diagram that Fig. 3 shows according to an embodiment of the invention, the exemplary screen of the result of the specific search in source shows；

Fig. 4 shows when entity search is carried out according to embodiments of the invention, demonstrates and filters the schematic diagram that shows of exemplary screen sometimes being considered as necessity why；

Fig. 5 shows the schematic diagram that the exemplary screen of the entity search implemented according to embodiments of the invention shows；

Fig. 6 shows the schematic diagram that the exemplary screen of another entity search implemented according to embodiments of the invention shows；

Fig. 7 be a diagram that the schematic diagram that the exemplary screen presented integrating the result searched for implemented according to embodiments of the invention shows；

The schematic diagram that Fig. 8 A and Fig. 8 B be a diagram that according to an embodiment of the invention, the exemplary screen presented of facet entity search shows；

Fig. 9 shows according to an embodiment of the invention, for determining the target of Web search based on entity type and resolving the flow chart of exemplary method of its result；And

Figure 10 shows according to an embodiment of the invention, for determining the target of Web search based on entity type and resolving the flow chart of another exemplary method of its result.

Detailed description of the invention

Subject of the present invention is specifically described to meet legal requirement in this article.But, description itself is not intended to limit the scope of this patent.On the contrary, the present inventor has contemplated that theme required for protection is likely to combine the technology of other present or future and is otherwise embodied as, to include different steps or the combination of the step similar with step described in the document.In addition, although term " step " and/or " square " can be used for the different unit of the method that hint is utilized in this article, but described term is not construed as among various steps herein disclosed or between implicit any specific order, unless and during except explicitly describing the order of single step.

The various aspects of technology described herein are generally directed to be particularly useful for integrating the advantage of vertical search and the advantage of Web search to utilize entity type to characterize the system of search experience, method and the computer-readable storage medium providing abundant.According to embodiments of the invention, " entity " is the description of certain real-world objects or project.It is to say, entity is different from the expression of the real world concept of Web document.It not that Web document is matched with appearance term in the search query, embodiments of the invention manage to make entity or real-world item (tangible or invisible) to match with inquiry, so that more rich search experience can be provided, as described more fully below.The entity sharing predicable can be grouped into many entity types.

According to embodiments of the invention, search inquiry is received and entity type is determined for described inquiry.Then Web search for described inquiry is performed for the predetermined list being considered for fixed entity type is authoritative source.Although certain source is probably authority for described entity type, but it can also include the information for other entity type, and similarly, identifies the result relevant to multiple entity types.Therefore, the result of the specific search in source is filtered based on described entity type, it is provided that the filtered list of result, described result respectively be defined as relevant with the entity of the entity type that described inquiry is associated.Described filtered list then by with resolve entity list compared with to determine that, by the equivalent entities of the different identifing source searched, described equivalent entities is combined into single potential Search Results.The Search Results of described merging is then based on rank value and is sorted.The described rank value of the entity being assigned to merging is the total rank value calculated from the single rank value provided for the entity from described not homology.Described result be then presented to user at least partially.

Therefore, one embodiment of the present of invention is for one or more computer-readable storage mediums of storage computer-useable instructions, when described computer-useable instructions is used by one or more calculating equipment, the one or more calculating equipment is made to perform for determining the target of Web search based on entity type and resolving the method for result of described Web search.Described method includes receiving search inquiry and determining at least one entity type for the search inquiry having been received by.Described method farther includes to perform Web search for the search inquiry having been received by, and described Web search is restricted to have been directed towards multiple sources that at least one entity type described is identified.Further, the result that described method includes filtering described Web search is to create the filtered list of Search Results, and each Search Results in the described filtered list of Search Results is relevant with the entity of at least one entity type described.Further, described method includes merging by the equivalent entities of the different identifing source in the plurality of source to create the merging list of Search Results, and each Search Results in the described merging list of Search Results is relevant with the different entities of at least one entity type described.Finally, described method includes being ranked up the described merging list of Search Results based on rank value, wherein, the rank value of at least one being assigned in the Search Results in the described merging list of Search Results is to add up to rank value, described total rank value is to calculate from the single rank value being provided for such entity, described entity with in the plurality of source described in not homology at least one of Search Results in described at least one is associated.

In another embodiment, the method that the present invention is directed to be performed by the one or more calculating equipment including at least one processor, described method for determining the target of Web search and resolving the result of described Web search based on entity type.Described method includes: be associated with the search inquiry received by least one entity type；Search for the multiple predetermined Web source identified at least one entity type described to determine the list of Search Results；And carry out the described list of filtered search result about at least one entity type described to create the filtered list of Search Results.Each Search Results in the described filtered list of Search Results is relevant with the entity of at least one entity type described.Described method farther includes the described filtered list of Search Results compared with the list of entities of parsing, in order to determine by the equivalent entities of the different identifing source in the plurality of predetermined origin；And the merging list of Search Results it is defined as being created by the described equivalent entities of the described different identifing source in the plurality of predetermined origin by merging.Each Search Results in the described merging list of Search Results is relevant with the different entities of described entity type.Further, described method includes being ranked up the described merging list of Search Results based on rank value, wherein, distribute to the rank value of at least one in the Search Results in the described merging list of Search Results from for described Search Results described at least one entity being associated and total rank value that the single rank value that is provided is calculated, described single rank value is by least some of offer of not homology described in the plurality of predetermined origin.

In still another embodiment of the invention, the present invention is directed to the system of a kind of target for determining Web search based on entity type.Described system includes that the calculating equipment being associated with search engine, described calculating equipment have one or more processor and one or more computer-readable storage medium and the data storage coupled with described search engine.Described search engine is configured to receive search inquiry, at least one entity type is determined for the search inquiry having been received by, identify authority source associated plurality of with at least one entity type described, search for the authoritative source of the plurality of identification to determine the list of Search Results, and the described list of filtered search result is to create the filtered list of Search Results.Each Search Results in the described filtered list of Search Results is relevant with the entity of at least one entity type described.Described search engine is further configured to the described filtered list of Search Results compared with the list of entities of parsing, to determine by the equivalent entities of the different identifing source in the plurality of authority source, and it is defined as being created by the described equivalent entities of the described different identifing source in the plurality of authority source the merging list of Search Results by merging.Each Search Results in the described merging list of Search Results is relevant with the different entities of described entity type.Described search engine is further configured to be ranked up the described merging list of Search Results based on rank value, wherein, distribute in the Search Results in the described merging list of Search Results described at least one rank value from for described Search Results described at least one entity being associated and total rank value that the single rank value that provides is calculated, described single rank value is by least some of offer of not homology described in the plurality of authority source.

Having briefly described the summary of embodiments of the invention, the EXEMPLARY OPERATING ENVIRONMENT that embodiments of the invention can be carried out wherein is described below, in order to provide the general context of the various aspects for the present invention.In general manner with reference to figure and especially with reference first to Fig. 1, it is illustrated and is typically specified as calculating equipment 100 for implementing the exemplary operating environment of embodiments of the invention.One example of described calculating equipment 100 simply suitable computing environment, and it is not intended to the use to embodiments of the invention or any restriction of proposition of functional scope.Calculating equipment 100 is not construed as having any dependence relevant to any one in illustrated component or combination or requirement.

Embodiments of the invention can be described in the general context of computer code or machine usable instructions, and described computer code or machine readable instructions include that the such computer of such as program module performed by computer or other machine (such as personal digital assistant or other handheld device) can be used or the executable instruction of computer.Generally, the program module including routine, program, object, component, data structure etc. refers to perform particular task or implement the code of specific abstract data type.Embodiments of the invention can be put into practice in various systems configure, and the configuration of described system includes portable equipment, consumer electronics, general purpose computer, more special calculating equipment etc..Embodiments of the invention can also be put into practice in a distributed computing environment, and wherein task is performed by the remote processing devices linked by communication network.

With continued reference to Fig. 1, calculating equipment 100 includes bus 110, and it directly or indirectly couples following equipment: memorizer 112；One or more processors 114；One or more present component 116；One or more input/output (I/O) port 118；One or more I/O components 120；And illustrative power supply 122.It is the things (such as address bus, data/address bus or combinations thereof) of one or more bus that bus 110 expresses possibility.Although for the sake of clarity, the various squares of Fig. 1 are shown by lines, but it practice, these squares represent (being not necessarily reality) component of logic.Such as, such for such as display device component that presents can be considered I/O component by a people.Similarly, processor has memorizer.Present inventors have realized that this is the character of this area, and reaffirm that the figure of Fig. 1 is only the explanation to the exemplary calculating equipment that can be used in conjunction with one or more embodiments of the invention.Do not distinguish between the such as such classification of " work station ", " server ", " kneetop computer ", " portable equipment " etc., because being all contemplated in the range of Fig. 1 and being referred to as " calculating equipment ".

Calculating equipment 100 typically comprises various computer-readable media.Computer-readable media can be any available media that can be accessed by calculating equipment 100, and including in any method or the volatibility of technology implementation and non-volatile media, removable and nonremovable media, it is used for storing such as computer-readable instruction, data structure, program module or the such information of other data.Computer-readable media includes but not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical disk storage apparatus, cartridge, tape, disk storage device or other magnetic storage apparatus, maybe can be used to store desired information and other medium any that can be accessed by calculating equipment 100.Any combination in above-mentioned is also included in the range of computer-readable media.

Memorizer 112 includes volatibility and/or the computer storage media of nonvolatile memory form.Memorizer can be removable, non-removable or combinations thereof.Exemplary hardware device includes solid-state memory, hard disk drive, CD drive etc..Calculating equipment 100 includes one or more processor, and it reads data from the such as such various entities of memorizer 112 or I/O component 120.(one or more) present component 116 and data instruction are presented to user or miscellaneous equipment.The exemplary component that presents includes display device, speaker, print component, vibration component etc..

I/O port 118 allows calculating equipment 100 to be logically coupled to include, and the miscellaneous equipment of I/O component 120, some in described I/O component can be built.Illustrative component includes microphone, stick, game paddle, dish, scanning device, printer, wireless device etc..

As mentioned previously, embodiments of the invention are generally directed to be particularly useful for integrating the advantage of vertical search and the advantage of Web search to utilize entity type to characterize the system of search experience, method and the computer-readable storage medium providing abundant.Search inquiry is received and entity type is determined for described inquiry.Then Web search for described inquiry is that the predetermined list in source of authority is performed for being defined as fixed entity type.Although certain source is probably authority's for described entity type, but it can also include the information for other entity type, and similarly, identifies the result relevant to multiple entity types.Therefore, the result of the specific search in source is filtered based on described entity type, it is provided that the filtered list of result, each result be defined as relevant with the entity of the entity type that described inquiry is associated.Described filtered list then by with resolve entity list compared with to determine the equivalent entities identified by the different sources searched, described equivalent entities is combined into single potential Search Results.The Search Results of described merging is then based on rank value and is sorted.The described rank value of the entity distributing to merging is the total rank value calculated from the single rank value provided for the entity from described not homology.Described result be then presented to user at least partially.

Referring now to Fig. 2, it is provided that the block chart of the Exemplary computing systems 200 that diagram embodiments of the invention can be utilized wherein.Generally, calculating system 200 illustrates the Web search that wherein can implement to determine target based on the entity type being determined for input search inquiry and the environment resolving its result.In other component having been not shown, calculating system 200 generally includes user and calculates equipment 210, search engine 212 and data storage 214, all communicates with each other via network 216.Network 216 can include, but is not limited to one or more Local Area Network and/or wide area network (WAN).Such networked environment is common in office, the computer network of enterprise-wide, Intranet (intranet) and the Internet.Therefore, network 216 is not further described.

It should be understood that any amount of user calculates equipment and search engine can be utilized in the calculating system 200 in the range of embodiments of the invention.Each can include individual equipment/interface or in distributed environment multiple equipment/interfaces of cooperative operation.Such as, search engine 212 can include multiple equipment and/or the module being arranged in distributed environment, and it jointly provides the functional of search engine 212 described herein.It addition, in other component/module having been not shown may also be included in that calculating system 200.

In certain embodiments, one or more in illustrated component/module may be implemented as independent application.In other embodiments, one or more in illustrated component/module can calculate equipment 210 via user and be implemented as service based on the Internet, or are implemented as the module within search engine 212.It will be understood by those within the art that, component/module illustrated in Fig. 2 in nature and is exemplary on number, and should not be construed as limited to.It is desired functional that any amount of component/module can be utilized to realize in the range of embodiments of the invention.And, component/module may be located on any amount of search engine or user's calculating equipment.As just example, search engine 212 as individual server (as shown), the cluster of server or can be provided away from the one or more calculating equipment in residue component.

It should be understood that this and other layout described herein is only used as example and is set forth.Other layout and unit (such as, the packet etc. of machine, interface, function, order and function) can be additional to or replace shown those to be used, and some unit can be omitted entirely.And, the many unit being described herein as are functional entitys, and it may be implemented as discrete component or distributed element or combines other component and be carried out, and is carried out in any suitable combination and position.The various functions being described herein as being performed by one or more entities can be performed by hardware, firmware and/or software.Such as, various functions can be stored the processor of instruction in memory by execution and perform.

User calculates equipment 210 can include any kind of calculating equipment, such as, the calculating equipment 100 such as described with reference to Fig. 1.Generally, client computing device 210 includes browser 218 and display 220.Browser 218 is especially configured to reproduce search engine homepage (or other online login page), and is rendered in association with search engine results page with the display 220 of client computing device 210.Browser 218 is further configured to user's input of the request for various webpages (including search engine homepage) that receives, receive the search inquiry of user's input (generally via the user interface input presented on a display 220, and permit alphanumeric and/or text be input in the search box specified), and be configured to such as receive the content for presenting on a display 220 from search engine 212.It should be pointed out that, that functional can the application by any other that can reproduce web content being described herein as being performed by browser 218 performs.Any and all such variants and their any combination are contemplated in the range of embodiments of the invention.

Search engine 212 is configured to receive request that the component (browser 218 such as, being associated with client computing device 210) that it is associated from the equipment that calculates with user receives and responds described request.It is in the field of the invention it will be recognized that the present invention can implement with any amount of research tool.Such as, internet search engine or database search engine can utilize the present invention.These search engines are well known in the art, and many similar processes shared by commercially available engine, and these processes are not further described.

As shown, search engine 212 includes that inquire-receive component 222, entity type determine that component 224, authority source determine that component 226, search component 228, filter element 230, entity merge component 232, sequence component 234 and present component 236.Illustrated search engine 212 also has the access right to data storage 214.Data storage 214 is configured to store the information relevant with search inquiry, entity and authority source.In various embodiments, such information can include, but is not limited to search query log, entity type and correspondence entity index, be defined as relative to indexed entity type be authority source index or the list of other entity enumerated and resolve.In an embodiment, one or more in the project that data storage 214 is configured to for storing otherwise in association can search for.It will be understood by those within the art that and understand, can be configurable with the data information that store explicitly of storage 214, and any information relevant to search inquiry, entity type and corresponding entity and the source that can search for can be included.The scope of the embodiment that the content of such information and capacity are not intended to limit the present invention in any manner.And, although data storage 214 is illustrated as single independent component, but it can essentially be multiple storage device (such as data-base cluster), part therein can exist explicitly with search engine 212, client computing device 210, another external computing device (not shown) and/or their any combination.

The inquire-receive component 222 of search engine 212 is configured to the request received for presenting the Search Results meeting input search inquiry.Typically, such request is received via the browser (browser 218 such as, being associated with client computing device 210) being associated with client computing device.In an embodiment, search inquiry can also implicitly be called and such as by following action or received otherwise by inquire-receive component 222, described action such as user points to (such as, on screen, television set or in physical world) something；Mouse pointer is moved to icon/text；Phone is talked with someone；Send SMS, push away spy (tweet) or state renewal.Search inquiry is input in the traditional directory input area that screen shows it is to say, embodiments of the invention are not limited to user.

Entity type determines that component 224 is configured to determine at least one entity type relevant to the search inquiry received.The standard technique that inquiry understands can be used for being mapped to query intention one or more entity type.Such as, the static state for entity type can be obtained by the static content in the data relevant to each type with dynamic associated score.As just example, such content can include the non-structured Web page in the text in the data base of related entities, related entities and be limited to those Web page Web link structure and for determining the training grader that the feature of the inquiry string feature when with entity type and correspondent entity thereof matches.Entity type can also be from context as by the action of following user or implicitly determined otherwise, and described user action such as user points to (such as, on screen, on television set or in physical world) something；Mouse pointer is moved to icon/text；Phone is talked with someone；Send SMS, push away spy (tweet) or state renewal.Any and all such variants and their any combination are contemplated in the range of embodiments of the invention.

If entity type determines that component 224 determines that more than one entity type is relevant to input search inquiry, then the result of following entities search can be flattened into single result of page searching, and the entity of each type is staggered in some way/mixes.Alternatively, the interface for facet search can be presented, and wherein, user can make search narrow down to the entity type that (one or more) advise, effectively filters out the result of some in following entities search from last result.This is more fully described below relative to Fig. 8 A and Fig. 8 B below.Any and all such variants and their any combination are contemplated in the range of embodiments of the invention.

Authority source determines that component 226 is configured to identify that being previously determined to be for given entity type is authoritative multiple sources.Authority source is all sources crawled or the subset in source that can obtain from the structuring summary of the subscription for entity type.These sources can specific to regional market, but they can also be global.As an illustrative example, the film source specific to American market can include IMDB, ROVI/AMG, NETFLIX, HULU, YAHOO！REVIEWS, FLIXTER and METACRITIC.Among those, authority source can be identified as IMDB(i.e.,www.imdb.com), ROVI(i.e.,www.Allrovi.com) and NETFLIX(is i.e.,www.netflix.com).

The selection which source to be authoritative to can depend on many factors.Extreme at one, institute is active be all probably authoritative.Extreme at another, only one is authoritative.Generally, good authoritative source is by those sources of the result produced about the search of source particular Web.Such as, it is not likely to be the good candidate for authority source by the source of Web search engine index.And extraordinary authority source candidate be probably from other web site urls many to, there is abundant Anchor Text and there is a source of the abundant metadata about the Source Site by Web search engine index.

The inquiry of the given user with the intention matched with specific entity type, then search component 228 is configured to by using Web search engine (such as, well-known commercial search engine BING provided by the Microsoft of Redmond, Washington) Source Site of each authority is performed Web search, in order to perform the specific search in source for input search inquiry.Such as, search for and inquire about " in Space no one can hear you scream (nobody can hear your scream in space) " movie property mated, search component 228 can send inquiry according to BING used as described below:

To IMDB:

http://www.bing.com/search?q=site%3Awww.imdb.com+in+space+no+one+can+hear+you+scream；

To NETFLIX:

http://www.bing.com/search?q=site%3Awww.netflix.com+in+space+no+one+can+hear+you+scream；

To ROVI:

http://www.bing.com/search?q=site%3Awww.allrovi.com+in+space+no+one+can+hear+you+scream.Fig. 3 shows the schematic diagram that the exemplary screen of the specific search in this source shows 300.As expected, film " Alien (outman) " emerged, because in the abundant non-structured text that the citation (including that user inquires about in this example) of this film is comprised on ALLROVI website, or because the hyperlink pointing to ROVI " Alien (outman) " webpage uses this citation as Anchor Text.Similar search can be by similar HTTP POST request or be issued to any main search engine via the API of the replacement emerged by Web search engine.The result of source particular Web search is the list of document, and the Web search engine that described document is queried is considered the most relevant to described inquiry.Depending on that what information can be provided by Web search engine, the inside associated score of described engine can also be for each document enumerated and is returned.

Filter element 230 is configured to filter each list of the document returned by the source particular Web search of authority, and only retains URL relevant to entity type interested potentially.Fig. 4 show be illustrated as what filter sometimes be considered as necessity exemplary screen show 400 schematic diagram.In example illustrated, the website, source of authority comprises many pages, and it is not directly related with entity type (in this example, entity type is " film ").Any one in multiple methods can be used for filtering out the incoherent page.Hereafter intend the restrictive embodiment illustrative rather than the present invention.Regular expression URL pattern manually can be determined for each entity type and authority source.To keep unfiltered with the source particular search result of described pattern match, and will be filtered with the unmatched URL of described pattern.In addition, described pattern can specify where the internal entity thought being derived from oneself is expressed, it can be used in the presenting of source identifying entity in entity merging/analyzing step, and described entity merging/analyzing step reference entity merges component 232 and is more fully described below below.Exemplary patterns for IMDB and NETFLIX includes:

IMDB pattern: www.imdb.com/title/{ID starting with tt}；

NETFLIX pattern: www.netflix.com/Movie/{string}/{ numeric id}。

These URL pattern can also be extracted from the authoritative website, source relevant to given entity type in the case of given one group of exemplary document.Entity search engine 212 can emerge API, website, source and summary on the contrary and submit URL pattern to by this API.Or, Web page can use predetermined standard to be embedded on those pages with the dependency of particular entity type together with inside sources ID.Any and all such variants and their any combination are contemplated in the range of embodiments of the invention.

If internal data source ID is can not to obtain from the page emerged the specific search in source, then entity resolution can be used for matching those page results with the entity from data source.Merge component 232 below with reference to entity entity resolution is described more fully.Described by here, this additional application of entity resolution can provide the mapping between the URL emerged and internal ID, thus enable and merged, by entity, the filtration that component 232 is carried out.

It is active that entity merging component 232 is configured to run the institute for entity type by entity resolution pipeline, and is configured to the result of entity resolution compared with the particular result of filtered source.The collection of the given entity from multiple sources, entity resolution represents the process of two steps on high-level, equivalent entities between its parsing source the and then set merging of equivalent entities becomes single, the more rich expression of basis entity.Initially, only entity resolution is necessary.The merger of final step to(for) method described herein after a while is needs, and wherein, the expression of the single merger of the entity that inquiry is relevant is emerged by entity search, as described more fully below.Parsing and merger utilize feature or the attribute of entity.Example for entity type " film " includes title, issues time, (one or more) director, cast, projection time, studio, (one or more) school etc..Name, birthday/age, sex, occupation, geographical position, home address, telephone number and other personally identifiable information can be included for entity type " people ", feature or attribute.Entity from the identical entity type of not homology is compared about their attribute, ultimately results in the set producing matching entities.These set to matching entities use merger, and produce the entity of prototype for each single attribute by merger and combination member entities therein.

The result of the result of entity resolution search specific with filtered source is combined.Because the specific search in source is run on source that run by entity resolution, solid data, so search engine 212 can be grouped by the document that retrieved of search, described document is known which are relevant entity (owing to filter element 230) and has been carried out mating (by using the result of entity resolution) by all the expression known to other with identical entity.By this way, the document relevant from entity identical, the specific search in filtered source is grouped by search engine 212.From source, specific search chain receives entity resolution result by using source internal entity ID to be done, as described above.

Each authority source is assigned quality coefficient Q.The total number of the entity in the source that described coefficient can be emerged during being based on Web search.The result of this step is the set of the URL relevant to the inquiry that user submits to, the identical entity of the entity type (as by represented by input search inquiry) that its expression is interested.Such as, two following URL can include in these results sets for Batman (batman) 1989 film:

www.imdb.com/title/tt0096895；

www.netflix.com/Movie/Batman/287290。

Sequence component 234 is configured to be equal to extraction in searching for from entity resolution and filtered source particular Web or the set of identical entity is ranked up.If drying method can be followed produces such ranking.Two main sources in signal are extracted for ranking.First source be the document from the search of source particular Web ranking (and the associated score returned by Web search engine alternatively, if any).In the case of not having associated score, these Search Results produce ranking (such as, first document may be allocated ranking one, and second is ranking two etc.) for each filtered document.If available, associated score can be assigned to the document returned.For can come from the inquiry string application normal data library searching in the data source being input in entity resolution in second source of signal carrying out the relative set of equivalent entities using in ranking.Such as, inquiry string can be resolvable to word, and these keywords mate with entity attributes.Depend on for entity search engine can data, the extra source of signal is also likely to be available.Grading such fashion information for example, it may be possible to there is such as user, described user grading is used in for recommending to carry out collaborative filtering.Another example can be the historic click-through rate data of the merger entity emerged by entity search engine about the past.Along similar circuit, toolbar data from authority's website can be used in general Web search ranking, but entity search is probably more relevant, making by they being included in entity search as signal, they can receive than them for the higher weighting of weighting received by the contribution of they associated score specific for source.

Ranking or associated score combination step are then applied to the specific search rank in source, search for the set in other source specifically searching for associated score, search associated score based on attribute and coherent signal.Substantially, such step is by the multiple associated score combination from these many individually rankings, as the feature of the first ranking for the entity being grouped.Then, the result of this step is the ranking about different entities, and described different entity is grouped by the specific search in single source by entity resolution.

Present component 236 and be configured to present ranked list at least some of of the entity relevant to the inquiry that user submits to.Merger according to entity resolution can be used for merger from across different crawl/index/attribute individually represented of the entity in source subscribed to.Illustration/image, metadata, entity behavior and the even specific advertisement of entity can be emerged with natural by entity search system, as illustrated in the screen of Fig. 7 described more fully below shows in 700 by the way of consistent.

It addition, search engine 212 can utilize facet search to advise that their Search Results can refine the most reduced or further to user.Such as, if multiple entity type is emerged in the result of page searching integrated, then user can be provided a kind of right to choose to cancel Selective type (such as, retain restaurant from result but remove coffee-house).Facet can be present in single entity type with more fine-grained rank.Such as, if presenting film result to user, then search engine 212 can allow a user to specify school (such as, science fiction) and publication year scope (such as, the film of distribution after 2000).Being illustrated in the screen of Fig. 8 A and Fig. 8 B shows for presenting an embodiment of facet search, they are more fully described below below.

Turn now to Fig. 5, thus it is shown that exemplary screen shows 500, which illustrates according to an embodiment of the invention, for the entity search example inputting search inquiry " James Bond (James Bond) ".Initially, entity type has been previously identified as " film ".Therefore, in this embodiment illustrated, it can be assumed that user is expected to James Bond (James Bond) franchise operation (such as, on NETFLIX) looks for film, but does not knows Film Title.Even if input inquiry character string is (i.e., " James Bond ") it is not included in any one in the Film Title finally emerged (and other attribute for the desired film on NETFLIX may be not belonging to), film as result is also emerged, because described inquiry string can occur in the Anchor Text being linked to NETFLIX, or occur in potentially and be resolved for NETFLIX and by another source of search engine index (such as, with comment " Diamonds Are Forever is my favorite James Bond movie (eternal diamond be me favorite James Bond film) " blog text) in.And, to search for by performing integral entity according to an embodiment of the invention, search engine (such as, the search engine 212 of Fig. 2) can emerge the metadata about relevant entity.Notice, be the multiple selectable icon representing authority source and/or entity behavior (such as, buy tickets, flow, taxi etc.) on the right of the movie property emerged.User can select such as NETFLIX icon, and be navigated to the NETFLIX version of the James Bond film of correspondence according to exemplary intention.

With reference to Fig. 6, thus it is shown that exemplary screen shows 600, which illustrates another entity search example according to an embodiment of the invention, it is for input search inquiry " Neo and Trinity " specifically.Again, entity type has been previously identified as " film ".As illustrated, user is just finding one in the film in Matrix trilogy, but forgets the name of film, only remembers the name of two high priests.By including that cast and personage, reviewer and fan that they play comment on, integrate the specific search in source on the website of plot etc., the search engine performing entity search according to an embodiment of the invention can retrieve desired entity.Notice, in the summary of the film that character name " Neo " and " Trinity " unlikely source being already present in and search for traditionally are associated.

With reference to Fig. 7, it is illustrated that show that 700 are illustrated according to the exemplary screen presented integrating the result searched for that embodiments of the invention are implemented.In example illustrated, user's inquiry is to quote from " there for the film from film Highlander Can be only one (can only have one) ".Notice that the merger result of the specific search of entity resolution and source is shown the most in a uniform matter.Each movie property is shown by illustration, title and metadata.The accordant interface of the icon that the vertical stacking at right-hand side is passed through in the different source providing the data of entity behavior (such as rent, flow, reads comment, purchase etc.) is emerged.It is also noted that only the document in desired entity type is emerged (owing to the specific search in source and filtration), and how each different entity only exists a result (owing to entity resolution and merging).

With reference to Fig. 8 A and Fig. 8 B, it is illustrated that the screen of facet entity search shows and is illustrated according to an embodiment of the invention.For initial ranging (Fig. 8 A), the entity of multiple types potentially maps to input search inquiry (that is, " Superman (superman) ").In example illustrated, " Superman " can refer to any one in entertainment entity type " film (Movies) ", " TV series (TV Series) " or " song (Songs) ".In the screen of the diagram of Fig. 8 A shows, user can select facet and make the characteristic of shown entity reduce by doing that.Such as, Fig. 8 B illustrates user's selection of entity type " TV series (TV Series) ".

Turning now to Fig. 9, show that the flow chart of exemplary method 900 according to an embodiment of the invention is illustrated, described method is for determining the target of Web search based on entity type and resolving its result.Initially, as indicated by square 910, such as, received search inquiry by the inquire-receive component 222 of the search engine 212 of Fig. 2.As indicated by square 912, at least one entity type is determined for the search inquiry having been received by.This can be such as by utilizing the entity type of the search engine 212 of Fig. 2 to determine that component 224 completes.Web search is performed for the search inquiry having been received by, (such as, by utilizing the search component 228 of Fig. 2) as indicated by square 914.Web search is restricted to multiple source, and it is authoritative that the plurality of source has been previously identified as at least one entity type described determined by the search inquiry having been received by.As indicated by square 916, the result of Web search is filtered to create the filtered list (such as, by utilizing the filter element 230 of Fig. 2) of Search Results.Each Search Results in the filtered list of Search Results determine with for the search inquiry having been received by described in the entity of at least one entity type relevant.The equivalent entities identified by the different source in the plurality of source is merged (as indicated by square 918), in order to create the merging list of Search Results.This can be such as by utilizing the entity merging component 232 of Fig. 2 to complete.The each Search Results merged in list of Search Results determine from for the search inquiry having been received by described in the different entity of at least one entity type relevant.As indicated by square 920, the described merging list of Search Results is such as by utilizing the sequence component 234 of Fig. 2 to be sorted for presenting based on rank value.Being assigned to the rank value at the Search Results merged in list of the Search Results representing equivalent entities is the total rank value calculated from the single rank value provided at least one of entity from the described different source the plurality of source.

With reference to Figure 10, the flow chart of display another exemplary method 1000 according to an embodiment of the invention is illustrated, and described method 1000 is for determining the target of Web search based on entity type and resolving its result.Initially, as indicated by square 1010, such as by utilizing the entity type of the search engine 212 of Fig. 2 to determine, component 224 is associated at least one entity type with the search inquiry received.As indicated by square 1012, it is identified as authoritative multiple predetermined Web source at least one entity type described and is searched to determine the list (such as, by utilizing the search component 228 of Fig. 2) of Search Results.As indicated by square 1014, the result of described search is filtered to create the filtered list (such as, by utilizing the filter element 230 of Fig. 2) of Search Results about at least one entity type described.Each Search Results in the filtered list of Search Results determine with for received search inquiry described in the entity of at least one entity type relevant.As indicated by square 1016, the filtered list of Search Results is then by compared with the list of entities resolved, in order to determine the equivalent entities identified by the different source in the plurality of predetermined source.It is defined as by the equivalent entities of the different identifing source in the plurality of source merged (as indicated by square 1018), in order to create the merging list of Search Results.This can be such as by utilizing the entity merging component 232 of Fig. 2 to complete.The different entity of at least one entity type described in determining from for received search inquiry at each Search Results merged in list of Search Results is relevant.As indicated by square 1020, the merging list of Search Results is such as by utilizing the sequence component 234 of Fig. 2 to be sorted for presenting based on rank value.Being assigned to the rank value at the Search Results merged in list of the Search Results representing equivalent entities is the total rank value calculated from the single rank value provided at least one of entity from the described different source the plurality of source.

It is appreciated that and The embodiment provides the advantage of advantage and Web search for integrating vertical search to utilize entity type to characterize the system and method providing abundant search experience.The present invention is described already in connection with specific embodiment, and these specific embodiments are intended to be illustrative and not restrictive in all respects.In the case of without departing from its scope, the those of ordinary skill in the field belonging to the present invention be will be apparent from by the embodiment of replacement.

Although the present invention is vulnerable to the impact of the structure of various amendment and replacement, but the embodiment of some diagram of the present invention is illustrated in the drawings and is described in detail above.It will be appreciated, however, that there is not the intention limiting the invention to particular forms disclosed, but just the opposite, it is intended that contain all modifications, the structure of replacement and the equivalent being within the spirit and scope of the invention.

It will be understood by those within the art that, the order of the step shown in the method 1000 of the method 900 and Figure 10 of Fig. 9 is not meant to limit by any way the scope of the present invention, and it practice, described step can occur with various different orders in an embodiment of the present invention.Any and all of such variants and their any combination are contemplated in the range of embodiments of the invention.

Claims (17)

1. for determining the target of Web search based on entity type and resolving the method for result for described Web search, described Method includes:

Receive search inquiry；

Determining at least one entity type relevant to the described search inquiry received, wherein said entity type represents by least one At least one predicable that entity and other entity one or more are shared；

Identify that multiple authority source, described authority source include for being confirmed as at least one entity relevant to received search inquiry Type is the Web source of authority；

The Web search of each in authority source is performed to identify and looking into received search for received search inquiry Ask relevant Search Results；

For each in authority source, filter based on the entity type that the search inquiry being confirmed as to received is relevant and receive from this source Search Results in case create Search Results filtered list, each search in the described filtered list of Search Results Result all relevant to received search inquiry and with and be confirmed as relevant at least one with received search inquiry The entity that entity type is associated is relevant；

Equivalent entities is merged to generate the set of equivalent entities, the set of each equivalent entities across each of filtered list Relevant to received search inquiry and with and be confirmed as at least one entity class relevant with received search inquiry The different entity that type is associated is relevant；And

Based on rank value, the set of described equivalent entities is ranked up, wherein, distributes at least one of set of described equivalent entities The rank value of set is to add up to rank value, described total rank value be from by authority source at least some of for described at least one The independent rank value that gathering relevant described entity provides calculates.

Method the most according to claim 1, wherein, filters the result of described Web search to create the described of Search Results Filtered list includes at least partially through utilizing URL pattern to filter the described knot of described Web search Really.

Method the most according to claim 1, wherein, merges equivalent entities so that the set generating equivalent entities includes searching for The described filtered list of result is compared with the list of entities of parsing, in order to determine by the different power in the plurality of authority source The described equivalent entities of prestige identifing source.

Method the most according to claim 1, wherein, described method farther includes to present described ranked equivalent entities That gathers is at least some of.

Method the most according to claim 4, wherein, the portion being presented described in the set with described ranked equivalent entities The information that at least one Search Results in Fen is associated is that the multiple authority sources from the plurality of authority source are derived.

Method the most according to claim 5, wherein, at least of the multiple authority sources described in from the plurality of authority source The described information of individual middle derivation is not presented explicitly with at least one Search Results described.

Method the most according to claim 5, wherein, it is described that described method farther includes to present in the plurality of authority source Multiple authority sources at least one of instruction, the described information being associated with at least one Search Results described is to be derived from 's.

Method the most according to claim 7, wherein, the described instruction in the described multiple authority sources in the plurality of authority source Be selectable at least partially.

9. one kind by include at least one processor one or more calculating equipment perform for determining that Web searches based on entity type The target of rope and resolve the method for result of described Web search, described method includes:

Receive search inquiry；

At least one entity type that the search inquiry that determines and received is relevant, wherein said entity type represents by least one entity At least one predicable shared with other entities one or more；

Identify that multiple Web authority source, described authority source include at least one entity class for being associated with the search inquiry received Type is the Web source of authority, described authority source include with and at least one entity type phase of being associated of the search inquiry that received One or more in the subset in the structuring summary of subscription closed or the Web source crawled；

Using the search inquiry received, each in search authority source is to determine the search knot relevant to the search inquiry received Really；

For each in authority source, based at least one the entity type filtered search result being associated with the search inquiry received To create the filtered list of Search Results, each Search Results in the described filtered list of Search Results and institute The search inquiry received relevant and with and the entity that is associated of at least one entity type described relevant；

Use the one or more calculating equipment by each in the described filtered list of Search Results and the list of entities resolved Compare, in order to each in the described filtered list of Search Results determines equivalent entities；

Search Results is created by merging each the described equivalent entities determined in the described filtered list of Search Results Merge list, each Search Results in the described merging list of Search Results relevant to received search inquiry and with The different entity that at least one entity type being associated with received search inquiry is associated is relevant；And

Based on rank value, the described merging list of Search Results is ranked up, wherein, distributes in the described merging list of Search Results Search Results in the rank value of at least one be at least some of for relevant with at least one of Search Results from by authority source Described entity provide independent rank value calculate total rank value.

Method the most according to claim 9, wherein, based at least one entity class being associated with the search inquiry received Type filtered search result is so that the described filtered list creating Search Results includes at least partially through utilizing unified resource to determine Position symbol pattern carrys out the list of filtered search result.

11. methods according to claim 9, farther include to present at least one of the described ranked list of Search Results Point.

12. methods according to claim 11, wherein, are presented described in the described ranked list with Search Results The information that at least one Search Results in part is associated is that the multiple sources from multiple predetermined origin are derived.

13. methods according to claim 12, wherein, at least one of the multiple sources described in from the plurality of predetermined origin The described information of middle derivation is not presented.

14. methods according to claim 12, the described multiple sources farther including to present in the plurality of predetermined origin are extremely At least part of instruction, the described information being associated with at least one Search Results described is derived from.

15. methods according to claim 14, wherein, the described instruction in the described multiple sources in the plurality of predetermined origin It is selectable at least partially.

16. 1 kinds for determining the target of Web search based on entity type and resolving the equipment of result of described Web search, institute The equipment of stating includes:

For receiving the device of search inquiry；

For determining the device of at least one entity type relevant to the described search inquiry received, wherein said entity type represents At least one predicable shared by least one entity and other entity one or more；

For identify multiple authority source device, described authority source include for be confirmed as relevant to received search inquiry extremely A few entity type is the Web source of authority；

For performing the Web search of each in authority source for received search inquiry to identify and searching with received The device of the Search Results that rope inquiry is relevant；

For for each in authority source, the entity type relevant based on the search inquiry being confirmed as to received filters from this source The Search Results received is to create the device of the filtered list of Search Results, in the described filtered list of Search Results Each Search Results relevant to received search inquiry and with and be confirmed as with received search inquiry be correlated with The entity that is associated of at least one entity type relevant；

For merging equivalent entities across each of filtered list to generate the device of the set of equivalent entities, each equivalent is real The set of body all relevant to received search inquiry and with and be confirmed as with received search inquiry relevant at least The different entity that one entity type is associated is relevant；And

For the device set of described equivalent entities being ranked up based on rank value, wherein, the set of described equivalent entities is distributed to Rank value of at least one set be to add up to rank value, described total rank value be from by authority source at least partially for institute The independent rank value that stating the relevant described entity of at least one set provides calculates.

17. 1 kinds for determining the target of Web search based on entity type and resolving the equipment of result of described Web search, institute The equipment of stating includes:

For receiving the device of search inquiry；

For the device of relevant at least one entity type of the search inquiry determined to received, wherein said entity type represents by extremely At least one predicable that a few entity and other entity one or more are shared；

For identifying the device in multiple Web authority source, described authority source includes for being associated at least with the search inquiry received One entity type is the Web source of authority, described authority source include with and the search inquiry that received be associated at least one One or more in the structuring summary of the subscription that entity type is relevant or the subset in Web source that crawls；

The search inquiry received for use, each in search authority source is to determine search relevant to the search inquiry received The device of hitch fruit；

For for each in authority source, based at least one the entity type filtered search being associated with the search inquiry received Result is to create the device of the filtered list of Search Results, each search in the described filtered list of Search Results Result all relevant with received search inquiry and with and the entity that is associated of at least one entity type described relevant；

For using the one or more calculating equipment by each in the described filtered list of Search Results and the entity resolved List compares, in order to each in the described filtered list of Search Results determines the device of equivalent entities；

For creating search knot by merging each described equivalent entities determined in the described filtered list of Search Results The device merging list of fruit, each Search Results in the described merging list of Search Results and received search inquiry Relevant and from and the different entity that is associated of at least one entity type of being associated of received search inquiry relevant；With And

For the device described merging list of Search Results being ranked up based on rank value, wherein, distribute to the described of Search Results Merge the rank value of at least one in the Search Results in list be from by authority source at least some of for Search Results to The total rank value that the independent rank value that a few relevant described entity provides calculates.